I will add I’m sad to see Anthropic go anti-community on their Claude subscription usage but since Codex works in pi then using pi as the underlying harness is probably the best thing to do. I did use open source codex rust in cleon to easily bring that into python directly. Big thanks and props to OpenAI for building their coding harness in rust and doing it open source.
Yeah I use it myself when I need to quickly debug something in Jupyter because I can just pip install cleon anywhere and then do @ question to get codex involved.
I took Jeremy’s solveit course and built this in homage to the concept of AI agent within Jupyter.
Keen to collaborate with anyone if they want to take this paradigm further.
I built something similar with just plain cli agent harnesses for Jupyter a while back.
It supports codex subscriptions and pi, (used to support Claude subs, might still be okay since I didn’t modify the system prompt).
Has some bugs and needs some work but getting help and code changes inline in Jupyter is way better than copy pasta hard to select text from cells and cell output all day.
Cal Newports Slow Productivity has a lot to say about this topic.
Creativity comes in bursts and can’t be scheduled. Happiness and health move with the seasons. Treating humans as divisible units of 1 hour blocks of factory farmed ROI will never yield amazing results.
It’s sad to see the more technology and automation we invent the more we become slaves to the cult of pseudo-productivity, virtue signalling hours at work in place of meaningful output or results.
I agree with this too, I have read a lot of academic stuff with - em dashes and it reads like the author doesn't know how to write and is just translating spoken word to text.
When you are talking, an aside can make a lot of sense because you are thinking and speaking in real time. When you write you have the luxury of time to reformulate your words more precisely. Em dashes are best kept for prose that mimics speech rather than constructing logical text.
It's no coincidence that em dashes are rare in legal texts because they are too imprecise. Where as semicolons are extremely common in legal texts.
The S in semicolon stands for S-Tier.
Maybe the E in em dash stands for E-Tier?
I 100% agree. I hear a lot of pushback from em dash lovers on the semicolon; if you can learn to love an em dash, you can learn to use and love a semicolon.
Remember, meaning is based on common usage, so now em dash is slop-nonymous, semicolons can take on a more casual vibe.
For example:
I love pizza — it's my comfort food.
Can just become:
I love pizza; it's my comfort food.
For asides:
I love pizza — especially pepperoni.
Can just become:
I love pizza (especially pepperoni).
Conformant OpenGL 4.6, OpenGL ES 3.2, and OpenCL 3.0? Check.
Conformant Vulkan 1.4? Check.
Proton gaming? Check.
Bravo!
Just think of the value that will have downstream for game devs or others who wish to target macOS. Apple really should do more to publicly acknowledge this kind of work and donate to these kinds of projects since they are beneficiaries.
Theres a chance here for them to open up further support for Linux across the hardware, software chain (Swift), containerization (finally) and maybe even allow app store virtualization on iOS (think running a fast linux compliant container on your phone via an intent or shortcut).
If not, perhaps DHH's Omarchy will be the Barbarians (with great hair) at the Gates which could turn the tide back away from Apple dev hardware dominance. If we could only get something like that on our phones without being blocked at every turn by "big app store".
Great article. Lots of similarities to other terms floating around like forward deployed engineer and full stack builder. This is obvious if you ask ChatGPT to list all the professions over the years which have been combined and subsumed during technical innovation.
I think the most obvious recent event would have been the personal computer. How many office jobs that existed pre-computer were combined once they became software?
Filing clerks, switchboard operators, darkroom technicians etc.
Totally agree, the FFI escape hatch and excellent tooling from rust, maturin pyo3 etc means so many python problems can just be solved with rust. Which begs the question, has anyone tried doing a greenthread implementation in rust? Maybe offload some of the dynamically evaled python code to a seperate process maybe with https://github.com/RustPython/RustPython
Something which would be immensely helpful for the community is to create a test suite of pathological problems with existing python concurrency patterns and libraries.
At that point it should just be a matter of time before the right implementation and PEP(s) can be iterated on that solves said problems while maximizing for devex.
For anyone interested in learning more about different concurrency models I can highly recommend Paul Butcher's: Seven Concurrency Models in Seven Weeks.
It’s actually really good at small bugs but you need to make sure it can understand what’s broken by giving it context. Logs, curls, screenshots etc. The biggest problem is when it goes on vision quests and does stuff you didn’t ask.
The meme with the car and duck are basically how AI agents like to “fix bugs”, sometimes they’re like oh I fixed the bug and changed the button which you never asked for. Lol. Adding a wheel and changing the car into a cat!
Agreed that behind every success is something which can be viewed retrospectively through a negative lens. The king of this is Uber.
Perhaps this is more a reflection of hyper-capitalism than tech itself but the is book is more about the ends and less about justifying the means. The word “hustle” is doing a lot of heavy lifting.
I imagine stories like gumroad and Basecamp are much better ones to idolize on a better way to create value in the world.
In some ways network effects and monopolies can be used synonymously and most would agree the later is a negative thing.
I think theres value in the books analysis of the early Cold Start strategies, which to me are reminiscent of the excellent blog post from Paul Graham https://paulgraham.com/startupideas.html
In fact you could read that and get a lot of the same value, particularly around "solving hard problems" and single-player / come for the tool, stay for the network.
But I do think his assessment that the whole thing is a paradox is true, and perhaps lends some credence to your point which is that, you kind of can't entirely engineer (or predict) success because the outcome is often different to how it started.
Paying for users (he calls it "subsidies") is mentioned but Chen talks about the fact that if you do it at the wrong time and you don't have hard side users or value and you're k-factor is too low, those users just churn. Theres a whole section dedicated to the strategic use of subsidies which have been used to great effect once you have a system of cyclic value generation.